Training to failure has an intuitive appeal that the evidence does not really support. The question is not whether it works, it is whether it is a good trade, and across a training week it usually is not.
The trade
What failure buys: a small and inconsistent increase in the growth stimulus from that set.
What it costs: substantially more fatigue, both local and systemic, and a longer recovery before that muscle can be trained productively again.
Since hypertrophy is driven largely by weekly volume, and failure reduces how much volume you can recover from, taking every set to failure often produces less growth over a month even though each individual set was harder. That is the core of it.
Where failure is genuinely useful
Isolation exercises. A set of lateral raises to failure costs little systemically. A set of squats to failure costs your next two days.
The last set of an exercise. A common and sensible compromise: stop short on the working sets, take the final one to or near failure.
Calibrating your sense of RIR. Genuinely valuable and under-used. Most people think they are at one rep in reserve when they are at four. Occasionally going to true failure recalibrates that.
When volume is limited. If you only have two sets, making them count matters more than the recovery cost.
Where it is a bad idea
Heavy compound lifts. Squats, deadlifts and heavy pressing. High systemic cost, technique degradation at the limit, and a genuine injury risk when form fails under load.
Early in a session. Failure on the first exercise compromises everything after it.
Every set, permanently. The route to accumulating fatigue faster than you clear it, which is the pattern in the signs of overtraining.
Judging how close you actually are
Bar speed. The most reliable cue available without equipment. When the rep slows noticeably despite you pushing as hard as you can, you are within a couple of reps of failure.
Form breakdown. When technique starts changing to complete the rep, you are there or past it.
Not how it feels. Perceived effort is a poor guide, and the error runs in one direction: people believe they are closer to failure than they are, which is why prescriptions of two RIR are frequently executed at four or five.
What to do instead
Take most working sets to roughly one to three reps in reserve, put the effort you saved into more total quality sets across the week, and let the last set of an exercise get close to the limit if you want it to.
That is more productive than a heroic session followed by three days of being unable to train the same muscle group. The weekly arithmetic is in how many sets per muscle per week.
On planning the fatigue rather than absorbing it, see what is a deload week.